Multilevel allometric growth equations improve accuracy of carbon monitoring during forest restoration

Abstract

Managing disturbed forests for climate mitigation and biodiversity requires monitoring the carbon (C) cycle consequences of replacing established exotic vegetation with native seedlings. Standard approaches rely on allometric growth equations with unexplored limitations for measuring C changes during restoration. Most plants lack species-specific allometric growth equations. Additionally, these equations may perform poorly for different growth forms, especially when applied to both mature trees and seedlings. To address these limitations, we generated and compared allometric growth equations for four woody species with different biogeographic origins and growth forms, including two high impact invasive species, Cupaniopisis anacardioides and Schinus terebinthifolia. By borrowing strength from sampling across species to reduce estimation error within species, Bayesian multilevel models generated more accurate estimates than either independent species-level models or generic equations, although improvements over independent species-level models were modest. Because errors systematically changed with plant size, especially for species with unusual growth forms, allometric growth equations from custom multilevel models generated higher baseline aboveground biomass estimates and lower post-restoration estimates, which has important implications for monitoring C consequences of invasive tree management.